A Robust and Practical Random Number Generator

In 2007 a novel ring oscillator based true-random number generator design (Rings Design) was introduced in [1]. The design was rigorously analyzed under a mathematical model and its performance characteristics were established. However, from a practical viewpoint the paper left much unexplored. In this paper, we focus on the practical aspects of the Rings Design. In particular, we consider transistor level effects such as phase interlock, narrow signal rejection, transmission line attenuation, and sampler bias and determine their implications on the earlier analysis framework. We make recommendations for avoiding pitfalls in real-life implementations. Furthermore, we present experimental results showing that changing operating conditions such as the power supply voltage or the operating temperature may affect the output quality when the signal is subsampled. Hence, an attacker may shift the operating point via a simple non-invasive influence and easily bias the TRNG output. Finally, we propose modifications to the design which significantly improves its robustness against attacks, alleviates implementation related problems, and simultaneously improves its area, throughput and power performance. Index Words – Oscillator rings, true random number generators, cryptography.